Abstract. It has been recently reported that convolutional neural networks (CNNs) show good performances in many image recognition tasks. They significantly outperform the previous approaches that are not based on neural networks particularly for object category recognition. These performances are arguably owing to their ability of discovering better image features for recognition tasks through learning, resulting in the acquisition of better internal representations of the inputs. However, in spite of the good performances, it remains an open question why CNNs work so well and/or how they can learn such good representations. In this study, we conjecture that the learned representation can be interpreted as category-level attributes that have good properties. We conducted several experiments by using the dataset AwA (Animals with Attributes) and a CNN trained for ILSVRC-2012 in a fully supervised setting to examine this conjecture. We report that there exist units in the CNN that can predict some of the 85 semantic attributes fairly accurately, along with a detailed observation that this is true only for visual attributes and not for non-visual ones. It is more natural to think that the CNN may discover not only semantic attributes but non-semantic ones (or ones that are difficult to represent as a word). To explore this possibility, we perform zero-shot learning by regarding the activation pattern of upper layers as attributes describing the categories. The result shows that it outperforms the state-of-the-art with a significant margin.

Visual short-term memory (VSTM) has been investigated with a change detection task. Recent studies suggested that there might be some representations in VSTM even when a change was not detected. However, this is discrepant with the previous studies that estimated the representation by change detection. In this study, we investigated the properties of the representation to be retained between two stimuli in a change detection task combined with the probe method so as to explore what causes the discrepancy.The interval between the test and the comparison stimuli and the timing of a positional cue at the location of change were manipulated.The results of three experiments suggested that, before the comparison stimulus presentation, the representations in VSTM were retained more than representations estimated by a normal change detection task, that they decayed with time, and that their availability decreased when the representations of the comparison stimulus were formed. From these results, we discussed a model of VSTM with attention.

The basic policy of bridge management is preventive maintenance which contains to execute bridge inspection, diagnosis, repair design, and records continuously. Deep learning which has been widely used in bridge inspection is not well examined in repair design because it is more complex and requires explainability. This study aims to propose a bridge repair decision model consisting of a damage grade assessment model and a damage progress decision model for repair design. The model trained by using periodic inspection data including images and texts achieved about 0.7 recall rate in bridge repair decision task.